Artificia lintelligence-assisted spatial stratification of tumor-infiltrating lymphocytes for predicting neoadjuvant therapy response and prognosis in triple-negative breast cancer
XIE Ting
CHENG Lingli
XIANG Lingyan
YAN Honglin
WANG Qianwen
YUAN Jingping
Abstract：Objective To evaluate the predictive value of tumor-infiltrating lymphocytes(TILs)in HE-stained slides of triple-negative breast cancer(TNBC)patients using artificial intelligence(AI)-assisted quantification before neoadjuvant therapy.Methods Clinical and pathological data from 123 TNBC patients who received neoadjuvant therapy were collected from three hospitals.Semi-automated annotation of HE-stained biopsy slides was performed to train a random forest model.Three spatially stratified TILs variables were constructed:intratumoral TILs(eTILs),stromal TILs(esTILs),and global TILs(etTILs),to assess the value of TILs from a spatial perspective.Results In the discovery cohort,esTILs demonstrated the best predictive performance for disease-free survival(DFS)(AUC=0.90),which was significantly superior to that of etTILs(AUC=0.78)and eTILs(AUC=0.61).Patients with high esTILs exhibited significantly prolonged DFS.For pathological complete response(pCR)prediction,etTILs emerged as the sole independent predictor in multivariate analysis(OR=16.58,95%CI:4.28-64.22,P＜0.01),and higher etTILs was associated with a higher likelihood of achieving pCR.These findings were robustly validated in an independent cohort.Conclusion This study reveal the differential predictive value of the spatial heterogeneity of TILs in TNBC:esTILs serves as an independent predictor for DFS,likely reflecting sustained immune surveillance function,while etTILs is specifically associated with pCR rate.
Keywords：Artificial intelligenceMachine learningRandom forestTriple-negative breast cancerTumor-infiltrating lymphocytesNeoadjuvant therapy response
Publication Date：2025-12-28
Online Publishing Date：2026-01-28(First online date of this platform, not the publication date of the document)
Pages：4( 1573-1576 )